A novel prognostic index—neutrophil times γ-glutamyl transpeptidase to lymphocyte ratio (NγLR) predicts outcome for patients with hepatocellular carcinoma

Clinical outcomes of patients with hepatocellular carcinoma (HCC) are highly variable. This study aims to identify and validate a simple, readily available, and objective prognostic index for the management of HCC. Data from 724 HCC patients undergoing curative resection were evaluated and randomly divided into two cohorts for building and validating the prognostic index. A best model, NγLR = (neutrophil count [109/L] × γ-glutamyl transpeptidase [U/L]) /(lymphocyte count [109/L] × U/L), was selected. An optimal cut-off value of 103.6 for NγLR stratified patients into high NγLR (>103.6) and low NγLR (≤103.6) groups. NγLR > 103.6 was closely associated with HCC malignant characteristics. Elevated NγLR predicted a worse overall survival (OS) and progression-free survival (PFS) for HCC patients and remained an independent predictor for both types of survival. Moreover, early recurrence rates in patients with NγLR > 103.6 were higher than that in patients with NγLR ≤ 103.6 (P < 0.0001). NγLR was an important independent predictor of survival for HCC patients and might be a new promising method to identify patients at different risks of early recurrence and survival after curative resection.


Materials and Methods
Patients. Between March 1993 and November 2010, 3,516 patients newly diagnosed with liver cancer at the Affiliated Hospital of Guilin Medical University (Guilin, People's Republic of China) were evaluated retrospectively. In the light of the included and excluded criteria, a total of 724 patients who underwent curative resection for HCC were eligible for this study (Fig. 1). The diagnostic criteria of HCC were based on clinical, serological, ultrasonography (US), computerized tomography (CT), magnetic resonance imaging (MRI), and pathologic examinations according to the Primary Liver Cancer Clinical Diagnosis and Staging Criteria (Ministry of Health, Beijing, China). Curative resection was defined as a complete resection of the tumour, a resection margin of at least 1 cm, no residual tumour based on histological examination, and no residual tumours or new lesions determined by two observations not less than 4 weeks apart 18 . Both intraoperative US and postoperative CT were Scientific RePORTS | 7: 9229 | DOI: 10.1038/s41598-017-09696-y used to confirm complete removal of HCC. The baseline information including demographic characteristics, hepatitis B surface antigen (HBsAg), hepatitis C virus antibody, use of nucleoside analogues (NAs), presence of liver cirrhosis, complete blood count, albumin, globulin, total bilirubin, direct bilirubin, alanine aminotransferase (ALT), AST, alpha-fetoprotein (AFP), γ-glutamyl transpeptidase (γ-GT), tumour characteristics, metastasis and recurrence were collected. All laboratory parameters used in the study were measured before curative resection. All methods were carried out in accordance with Affiliated Hospital of Guilin Medical University guidelines and regulations. This study was approved by the research ethics committee of Affiliated Hospital of Guilin Medical University and complied with The Declaration of Helsinki Principles. Informed consent was obtained from all patients.
Follow-up and postoperative treatment. All postoperative patients were followed with the regular monitoring of serum AFP concentrations and abdomen US every 2 months and chest radiography every 6 months during the first two postoperative years and at 3-6 month intervals thereafter. Further examinations, including CT and MRI, were performed if recurrence or metastasis was suspected. When recurrence or metastasis was confirmed, further treatment, i.e., a second surgical resection, radiofrequency ablation, transcatheter arterial chemoembolization, percutaneous ethanol injection or sorafenib treatment were suggested. The main clinical endpoint of this study was overall survival (OS), measured from the date of surgery to the date of death or the last follow-up. Progression-free survival (PFS) was calculated from the date of surgery to the date of recurrence, metastasis, death, or last follow-up. The cut-off value between early and late recurrence was set as 2 years. Statistical methods. All 724 patients were randomly assigned to either a training cohort for developing a new prognostic index or a validation cohort for evaluating the obtained prognostic index at an approximately 2:1 ratio. Variables were expressed as the mean ± SD unless otherwise stated. Categorical data were compared by the Pearson χ 2 test or the Fisher exact test, while continuous variables were assessed by Student's t test. OS and PFS were calculated using the Kaplan-Meier method and compared with the log-rank test. Univariate analysis was performed to identify significant prognostic factors by using the Cox regression model. The multivariate Cox proportional hazards regression model was performed to identify independent predictors by including all the variables demonstrated to be significant in the univariate analyses. All statistical analyses were performed with SPSS18.0 (SPSS Inc., Chicago, IL). Statistical significance was defined as P < 0.05.
Derivation of a prognostic model. To construct a best prognostic model, an exploratory formula was built by entering different sets of the independent factors into a regression model. The best model based on the preoperative peripheral blood neutrophil count, lymphocyte count, and serum γ-GT level was selected from the training cohort. The calculation formula for the novel prognostic model was as follows: NγLR = (neutrophil count [10 9 /L] × γ-GT [U/L])/(lymphocyte count [10 9 /L] × U/L). The diagnostic accuracy of the new index was estimated with the receiver operating characteristic (ROC) curve. Optimal cut-offs for NγLR were selected in terms of maximizing both the sensitivity and specificity. The ROC curves were compared between different prognostic models for HCC patients using the area under the ROC curves (AUC). Finally, the performance of the NγLR index was tested in the validation cohort.
Data availability. The datasets generated during and/or analysed during the current study are available from the corresponding author on reasonable request.

Results
Characteristics of the patients with HCC. Seven hundred and twenty-four HCC patients, with 463 in the training cohort and 261 in the validation cohort, were enrolled in this study. The clinical and laboratory characteristics of the HCC patients in both training and validation cohorts are shown in Table 1. Overall, the clinicopathologic characteristics between these two cohorts were very similar, except for smoking. The mean age of the patients in the training cohort was 50.14 ± 11.58 years and 49.56 ± 10.84 years in the validation cohort. In both training and validation cohorts, most patients had hepatitis B-related liver disease (85.3% and 82.8%, respectively) and cirrhosis (93.5% and 93.5%, respectively), and the numbers of HBsAg positive patients under NAs were 90 cases and 44 cases, respectively. Microvascular invasion was present in 17.7% of patients in the training cohort and 23.4% in the validation cohort. The median OS times were 47. 12

Comparisons of AUC between NγLR and other prognostic indices.
Several useful prognostic models resulted from previous research, such as SII and ALRI 16,17 . Therefore, we used the thresholds of these models to assess the prognosis of the HCC patients selected in the present study. The prediction abilities of the NγLR, SII, ALRI and conventional parameters, such as AFP, were compared. In the training cohort, the AUC for the NγLR was 0.758 (95% CI, 0.714-804), which was higher than those of SII (0.651, 95% CI, 0.602-0.703), ALRI (0.688, 95% CI, 0.641-0.738) and AFP (0.613, 95% CI, 0.562-0.665) ( Fig. 2A).
Early recurrence rate and further stratified NγLR in patients with HCC. The Kaplan-Meier curves also revealed that the NγLR > 103.6 group was associated with a higher early recurrence rate compared with the NγLR ≤ 103.6 group in the training (Fig. 3C, P < 0.0001) and validation cohorts (Fig. 3D, P = 0.007).

Prognostic values of NγLR in patients with early HCC (BCLC stage 0 and A).
In view of the prognostic values of NγLR in both training and validation groups, discriminative power of NγLR was further evaluated in early HCC (BCLC 0 + A) in greater detail. In the early-stage subgroup, NγLR > 103. 6 16; P < 0.0001; median PFS, 60.87 months; 95% CI, 55.09-66.64; P < 0.0001) in the training cohort ( Fig. 4A and C). In the validation cohort, NγLR also significantly correlated with OS and PFS in the BCLC 0 + A subgroup (P < 0.0001 and P < 0.0001, respectively, Fig. 4B and D). Correlation between NγLR and BCLC stage. BCLC stage is an important prognostic classification system for patients with HCC. Thus, we further analysed the relationship between the NγLR and the BCLC stage. Box plots of the NγLR in relation to the BCLC stage are presented in Fig. 5. In the training cohort, the severity of the BCLC stage was significantly positively correlated with a gradual increase in NγLR (r = 0.452, P < 0.001) (Fig. 5A). Furthermore, the results of the validation cohort (r = 0.415, P < 0.001) were the same as those obtained from the training cohort (Fig. 5B).

Discussion
In this study, we established a simple and evidence-based prognostic index, namely, the NγLR, which incorporates routinely available laboratory parameters to predict the risk of recurrence and poor survival in HCC patients undergoing curative resection. This prognostic index was both accurate and reproducible. Patients in different subgroups of NγLR levels had distinctly different prognoses after surgery. As a linear risk index, NγLR further categorized patients into low, medium, and high risk groups with significantly different survival rates. The performance of the NγLR in predicting the risk stratification for prognosis was validated in another independent validation cohort with similar accuracy.
In the current study, the NγLR consisted of three factors: neutrophil count, lymphocyte count, and serum γ-GT level. Neutrophils and lymphocytes are peripheral blood cells that are associated with systemic inflammatory response and immune response. Neutrophils are the key players in inflammatory disorders, while lymphocytes are reflective of the host immune response. Therefore, an increase of NγLR may suggest activation of the inflammatory status and an immune-suppressive response in patients. Elevations of NγLR are usually associated with neutrophilia, elevated γ-GT concentration or lymphopenia. Neutrophils can promote the growth, invasion, and migration of cancer cells by releasing proinflammatory, immunoregulatory, and angiogenic factors [19][20][21] . Elevated γ-GT was also associated with mortality from many causes, including liver disease, cancer, and diabetes, even when controlling for alcohol consumption and restriction to mild and non-drinkers 22 . Activation of cytotoxic T lymphocytes could induce tumour killing by released cytokines, such as tumour necrosis factor-alpha  Table 2. Correlation between the clinicopathologic variables and NγLR level in HCC patients (training cohort, n = 463 and validation cohort, n = 261). NγLR, neutrophil cell count times γ-glutamyl transpeptidase to lymphocyte count ratio; HBsAg, hepatitis B surface antigen; BCLC, barcelona-clinic liver cancer; AFP, alpha-fetoprotein; ALT, alanine aminotransferase; SII, systemic immune-inflammation index; ALRI, aspartate transaminase to lymphocyte ratio index; NAs, nucleoside analogues. and interferon-gamma. Meanwhile, endogenous T cells significantly delay malignant progression by responding to and infiltrating tumours 23 . Thus, it can be speculated that lymphopenia cannot effectively protect against the development of HCC by strengthening the body's immune response. Accordingly, all of these were adverse factors for HCC patients. The NγLR integration of these three factors can better reflect their impact on HCC. In addition, our findings show that high NγLR was positively correlated with larger and multiple tumour number, poor differentiation, vascular invasion, severe BCLC stage, and early recurrence. It is postulated that NγLR represents a systemic inflammatory response; the elevation of NγLR may play an important role in maintaining the malignant phenotype of HCC cells and persistent damage of liver inflammation via internal environment disorders, thereby promoting the recurrence and metastasis of HCC. Our observation is consistent with a gene expression analysis that indicated that HCC metastases and relapse might be promoted by a shift towards anti-inflammatory/ immune-suppressive responses 24 .
To date, AFP remains the most popular marker for the diagnosis and management of HCC. However, the sensitivity and specificity of AFP are limited; not all HCC patients secrete AFP. The sensitivity and specificity of AFP are only 61% and 81%, respectively, with a cut-off value of 20 ng/mL at the time of diagnosis 25 . Therefore, new markers are greatly needed to improve the ability to predict the prognosis of patients. In this study, we found that AUC for the NγLR was higher than that of AFP, and NγLR may be a novel prognostic marker for HCC patients. Moreover, our data showed that NγLR > 103.6 was also strongly correlated with the early recurrence of HCC. We know that the postoperative recurrence of HCC has two distinct mechanisms: the first is early recurrence via metastasis arising from undetected dissemination of tumour before surgery, and the other is late recurrence via de novo primary HCC in the liver remnant of dysplastic hepatocytes 26 . Postoperative recurrence, especially early recurrence, impacting the outcome of HCC is a very distressing situation for clinicians and patients. To date, no optimal marker has been identified. Although AFP is widely used in the postoperative monitoring of HCC, there   16 and ALRI 17 in HCC patients after surgery. The AUC of SII and ALRI to predict survival were lower than that of NγLR in this cohort; the same studies show inconsistent results, which may be associated with differences in the samples. Meanwhile, the prediction ability of NγLR was also higher than the conventional AFP. Therefore, NγLR had fairly good discriminatory power in stratifying patients with HCC into different prognostic groups. It is widely accepted that BCLC stage is the most popular staging system with the power of prognostic stratification and therapeutic allocation 27, 28 . In this study, when we further explored the relationship between the NγLR and the BCLC stage, we found that the NγLR gradually increased with the increasing severity of BCLC stages. Meanwhile, the elevation of NγLR was associated with increasing risks of death and recurrence of HCC, and the NγLR had a fairly good prognostic power in both training and validation cohorts. Our study also demonstrated that NγLR still had a strong prognostic significance in early HCC patients. On the whole, our data indicated that NγLR could serve as a powerful prognostic marker for patients with HCC. The predictive significance of the  NγLR in these subgroups is of great importance for clinicians to select the appropriate intervention after surgical resection to avoid unproven and futile treatment.
The NγLR was built in a cohort of patients with mainly HBV-related HCC. Effective antiviral therapy may potentially influence the prognosis of these patients. This finding is consistent with previously reported results that antiviral treatment is effective in a better survival 29 . However, antiviral therapy is prescribed in patients with a higher viral load and abnormal serum ALT levels, irrespective of cancer classification. This might explain why the predictive capability of NγLR may not be influenced by antiviral therapy.
The elevation of serum γ-GT concentration is strongly associated with heavy alcohol consumption. However, Ruhl et al. 22 reported that a high relative risk of liver disease mortality with elevated γ-GT was not influenced by alcohol consumption. Therefore, alcohol consumption does not diminish the accuracy of the NγLR index in predicting prognosis of HCC. Moreover, HCC patients with alcohol consumption often stop drinking alcohol under doctors' suggestions after being diagnosed with HCC.
We acknowledge that there are some limitations to our study. First, its retrospective nature is a potential limitation. Second, our study included HCC patients only from a single centre with a vast majority of patients having HBV infection. Although internally validated, whether our index can be generalized to different geographical areas remains to be determined. Third, the NγLR has been characterized in HCC patients undergoing curative resection; the prognostic ability of NγLR in palliative settings needs to be further explored in a future study.
In conclusion, we proposed a novel and simple prognostic index, the NγLR, including three readily available laboratory results for HCC patients after curative resection with a relative high degree of accuracy. The concept of the ratios among these three variables in the prediction of prognosis in patients with HCC is novel. NγLR can effectively identify the patients with HCC who are at the greatest risk of poor survival and early recurrence after surgery. It may be a useful tool for clinicians who manage postoperative patients with HCC. A prospective study to validate the NγLR index is being planned.